Scale beyond one machine. Run AI agents across your laptop, desktop, cloud servers, and team members' computers with SSH-powered auto-discovery and seamless orchestration.
Unlock the power of multiple machines working together
Split heavy tasks across machines. Run frontend agents on your laptop while backend services run on a powerful desktop or cloud instance.
Connect to AWS, GCP, or Azure instances for GPU-accelerated tasks, large-scale testing, or data-intensive operations without leaving your dashboard.
Share worker machines with team members. Everyone sees the same dashboard, same agents, same real-time terminals. Perfect for remote teams.
See CPU, memory, and network usage across all worker machines from one dashboard. Optimize agent placement based on real-time metrics.
All communication uses SSH with key-based authentication. No exposed ports, no security risks. Works with existing SSH configs and jump hosts.
Configure SSH hosts once. AI Maestro automatically discovers tmux sessions on all worker machines and displays them in your dashboard.
Manager/Worker architecture powered by SSH
┌─────────────────────────────────────────────────────────────────┐
│ MANAGER MACHINE (Your Laptop) │
│ │
│ ┌─────────────────────────────────────────────────────────┐ │
│ │ AI Maestro Dashboard │ │
│ │ http://localhost:23000 │ │
│ └─────────────────────────────────────────────────────────┘ │
│ │ │
│ │ SSH Auto-Discovery │
│ ▼ │
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ Local Agents │ │ Worker 1 │ │ Worker 2 │ │
│ │ (localhost) │ │ (desktop) │ │ (cloud) │ │
│ └──────────────┘ └──────────────┘ └──────────────┘ │
└─────────────────────────────────────────────────────────────────┘
│ │
│ │
┌────────────────┴─────┬───────────┴─────────────┐
│ │ │
▼ ▼ ▼
┌──────────────────┐ ┌──────────────────┐ ┌──────────────────┐
│ WORKER 1 │ │ WORKER 2 │ │ WORKER 3 │
│ Desktop (Linux) │ │ Cloud (AWS EC2) │ │ Team Machine │
│ │ │ │ │ │
│ tmux sessions: │ │ tmux sessions: │ │ tmux sessions: │
│ • backend-api │ │ • ml-training │ │ • integration │
│ • database │ │ • data-proc │ │ • e2e-tests │
│ • redis │ │ • gpu-tasks │ │ • perf-tests │
└──────────────────┘ └──────────────────┘ └──────────────────┘git clone https://github.com/\
23blocks-OS/ai-maestro.git
cd ai-maestro
./install.shInstall AI Maestro on your main machine (laptop/desktop). This becomes your manager.
# On each worker machine
git clone https://github.com/\
23blocks-OS/ai-maestro.git
cd ai-maestro
./install-worker.shMinimal installation on worker machines - just tmux and SSH.
# ~/.aimaestro/workers.yml
workers:
- name: desktop
host: desktop.local
- name: cloud
host: ec2-user@aws-instance
- name: team-server
host: team@company.comDefine workers once. Auto-discovery handles the rest.
All tmux sessions from all workers appear in your dashboard automatically.
Pro Tip: Workers can be anywhere - local network, cloud, behind VPN, or even team members' machines. As long as SSH works, AI Maestro works.
Common scenarios where distributed architecture shines
You have a laptop for mobility and a desktop for heavy workloads. Run lightweight agents on your laptop, but offload database migrations, large builds, or ML training to your desktop.
Laptop: UI agents, API testing
Desktop: Database, Redis, ML training
Control: All from laptop dashboard
Working from coffee shop with limited bandwidth? Keep heavy tasks running on your home office machine. Access them securely via SSH from anywhere.
Coffee Shop: Active development
Home Office: Background jobs, builds
Same workflow: Seamless switching
Share access to worker machines across your team. Everyone can monitor integration tests, access shared development databases, or collaborate on pair programming sessions.
Shared Workers: CI/CD, staging env
Personal Workers: Individual dev
One Dashboard: Unified team view
Keep fast feedback loops local, but leverage cloud GPUs or large instances for specialized tasks. Spin up/down cloud workers as needed while maintaining local control.
Local: Fast iteration cycles
AWS/GCP: GPU training, big data
Cost: Only pay when cloud is used
Power user capabilities
Access workers behind corporate firewalls or VPNs using SSH ProxyJump configuration.
# ~/.ssh/config
Host production-worker
HostName 10.0.1.50
ProxyJump bastion.company.com
User deployerFull SSH config support - custom ports, specific identity files, connection options.
# ~/.ssh/config
Host cloud-gpu
HostName gpu.example.com
Port 2222
IdentityFile ~/.ssh/gpu_key
StrictHostKeyChecking noSet CPU/memory limits per worker to prevent any single agent from consuming all resources.
# workers.yml
workers:
- name: desktop
host: desktop.local
limits:
cpu: 50%
memory: 8GBWorkers go offline? No problem. AI Maestro automatically reconnects when they're back online.
# Automatic behavior:
# - Retry every 30 seconds
# - Exponential backoff
# - Mark offline in dashboard
# - Reconnect when availableExport, transfer, clone, and backup your AI agents across any machine
See full Agent Management guide →
Export any agent to a portable .zip file. Includes configuration, memory database, message history, and git repository references.
Agent Settings → Export → Downloads .zip file
Drop .zip files into AI Maestro on any machine. Automatically restores configuration, memory, and messages. Offers to clone git repositories.
New Host → Import Agent → Select file → Restored!
Create copies of agents on the same machine or across hosts. Perfect for spinning up multiple similar agents or creating backups before experiments.
Export → Import on same machine → New agent clone
Keep backups of your critical agents. Restore from any point in time. Perfect for team onboarding or disaster recovery.
Weekly exports → Cloud storage → Instant recovery
Configuration
Registry settings
Memory DB
CozoDB database
Messages
Inbox & sent
Git Repos
Remote URLs
Moving to a new laptop? Export all agents, import on new machine, and you're back to work in minutes.
Share pre-configured agent setups with new team members. Everyone starts with the same optimized configurations.
Move agents between local machine and worker hosts. Optimize placement based on workload requirements.
Multi-computer mode is excellent for distributed workloads. Consider Docker when you need:
Complete dependency isolation without affecting host system
Identical environments across development, staging, and production
Install in 15 minutes. Works with existing SSH setup. Unlimited workers.